[Bioperl-l] new_fast methods
vecchi.b at gmail.com
Thu Feb 26 11:19:35 EST 2009
I have a suggestion. Wouldn't it be nice if NYTProf's results (which are
nicely formatted in html and are highly browsable) for scripts exercising
different modules could be somehow uploaded so that anyone wanting to do
some optimizations already knew where the bottlenecks are?
I'm sure that long-time users already know pretty much where to look, but
doing it systematically for the more important modules and uploading the
main results would be, IMO, a nice way to quickly detect where the biggest
improvements in speed could be accomplished. Profiling results would
obviously have to be linked to the script and input files used, and also the
hardware used to run them should be specified. Absolute times wouldn't have
much meaning, but relative times (ie, percentage time spent on each
module/subroutine, etc) would.
2009/2/26 Chris Fields <cjfields at illinois.edu>
> On Feb 26, 2009, at 7:52 AM, Jason Stajich wrote:
> FYI - I wrote some lightweight feature objects - there is a branch for it
>> (lightweight_feature_branch) - these had a pretty significant speedup.
>> A lot of the overhead with sequence/feature/location creation since there
>> are so many objects being created, so optimizing these features by using
>> arrays instead of hashes for the data structure seemed to provide a pretty
>> significant speedup as well. Ensembl uses a fast_new as well, right?
> I have been working on a Bio::Seq::RichSeqI that essentially does a
> fast_new (simple bless) and is a lazy parser, e.g. parses files for seq
> record start/end only, then does everything else (annotation, features, etc)
> on the fly. I'll probably be adding it to bioperl-dev sometime in the next
> few weeks.
> Jason, did you want to move the lightweight implementation off the branch
> to bioperl-dev? I worry about it getting ignored there.
> Bioperl-l mailing list
> Bioperl-l at lists.open-bio.org
More information about the Bioperl-l